Vertical Divider
Samsung Applies Machine Leaning to Smartphone RAM
Machine learning is used in smartphones to improve system performance and translations to voice commands and camera quality. Samsung revealed that its processing-in-memory (PIM) technology is coming to RAM used in smartphones, in addition to the tech landing in DRAM modules and high-bandwidth memory.
“Samsung’s LPDDR5-PIM mobile memory technology can provide independent AI capabilities without data center connectivity,” the company noted. “Simulation tests have shown that the LPDDR5-PIM can more than double performance while reducing energy usage by over 60% when used in applications such as voice recognition, translation, and chatbot (sic).”
It should result in faster, more efficient machine learning performance, but it’s unclear if PIM takes the form of an actual processor attached to the RAM module itself or is a suite of optimizations and algorithms. If it’s an actual chip, will third-party developers be able to access it for their own needs. It’s also unclear whether it can handle more advanced tasks like image processing and computer vision. Many smartphones already have chipsets that are capable of running advanced machine learning tasks, and virtually all flagship devices have dedicated AI silicon too. PIM-enabled RAM could theoretically be ideal for budget phones, allowing OEMs to perhaps go with cheaper silicon and let the “smart” RAM do some of the heavy lifting Samsung hasn’t revealed when we’ll actually see this RAM in smartphones, although it’s aiming to complete standardization of PIM technology in the first half of 2022.
Machine learning is used in smartphones to improve system performance and translations to voice commands and camera quality. Samsung revealed that its processing-in-memory (PIM) technology is coming to RAM used in smartphones, in addition to the tech landing in DRAM modules and high-bandwidth memory.
“Samsung’s LPDDR5-PIM mobile memory technology can provide independent AI capabilities without data center connectivity,” the company noted. “Simulation tests have shown that the LPDDR5-PIM can more than double performance while reducing energy usage by over 60% when used in applications such as voice recognition, translation, and chatbot (sic).”
It should result in faster, more efficient machine learning performance, but it’s unclear if PIM takes the form of an actual processor attached to the RAM module itself or is a suite of optimizations and algorithms. If it’s an actual chip, will third-party developers be able to access it for their own needs. It’s also unclear whether it can handle more advanced tasks like image processing and computer vision. Many smartphones already have chipsets that are capable of running advanced machine learning tasks, and virtually all flagship devices have dedicated AI silicon too. PIM-enabled RAM could theoretically be ideal for budget phones, allowing OEMs to perhaps go with cheaper silicon and let the “smart” RAM do some of the heavy lifting Samsung hasn’t revealed when we’ll actually see this RAM in smartphones, although it’s aiming to complete standardization of PIM technology in the first half of 2022.
Contact Us
|
Barry Young
|